Cooling is becoming the constraint: how data centre design is changing under pressure
Data centres have traditionally been designed with a focus on compute capacity, with cooling treated as a supporting system that can be scaled as needed. That assumption no longer holds. As demand for data processing continues to grow, driven by cloud services and artificial intelligence, cooling is becoming a primary constraint on how infrastructure is designed, operated and expanded. This shift is particularly relevant in Australia, where rapid growth in digital infrastructure is being balanced against constraints around energy availability, cost and sustainability expectations. As new facilities are developed and existing sites are expanded, operators are facing increasing pressure to optimise both performance and resource use. One of the clearest indicators of this change is the scale of energy required to run modern data centres. Cooling alone accounts for a significant portion of overall consumption, and total demand continues to rise as compute intensity increases. This is not just a question of efficiency. It is a question of whether existing approaches to cooling can keep pace with increasing thermal loads while remaining viable in a constrained energy environment. The result is a shift in how cooling is viewed. It is no longer a downstream engineering decision. It is a core determinant of performance, cost and scalability. This becomes more pronounced as rack densities increase. Traditional air-based systems remain effective at lower densities, but their limitations become more apparent as heat loads rise. Liquid cooling has emerged as a practical response. By using the thermal properties of liquids, systems can remove heat more efficiently and support higher-density infrastructure. However, this is not simply a technology shift. It requires a different approach to system design, integration and operation. In this environment, performance is no longer defined by capacity alone. It is increasingly defined by control. Cooling systems must operate in dynamic conditions, […]
